85% of Australia ’s population lives close to the coast and this is increasing. The coast is also extremely dynamic, as it is the last line of defense from forces such as extreme storms, waves, currents and sea-level rise. Understanding how coastlines change is therefore essential for adequate coastal planning and reducing the risk of natural disasters into the future.

This exciting project will make use of cutting-edge image segmentation algorithms from Meta/Facebook (https://segment-anything.com/) to develop a robust way of tracking coastal change from a smartphone (e.g. Android/iPhone) and other camera devices. The project will analyse a vast dataset of smartphone images collected all over the world as part of the CoastSnap community beach monitoring project (www.coastsnap.com), as well as the extensive network of coastal imaging cameras operated by the UNSW Water Research Laboratory. It will then evaluate the extent to which the Segment Anything Model can robustly identify shoreline positions over a wide variety of shorelines – and compare it to existing techniques.

School
Computer Science and Engineering
Research Area

Computer vision | Artificial intelligence | Coastal engineering | Image segmentation

This project is a unique opportunity to undertake applied artificial intelligence research to solve a critical problem related to coastal engineering. The UNSW Water Research Laboratory (WRL) and UNSW School of Computer Science and Engineering are leaders of their respective fields in coastal engineering and computer vision. WRL are the global developers of the CoastSnap community beach monitoring program which is operating in 27 countries worldwide. The two lead supervisors on this project (Dr. Harley and A/Prof Song) are prestigious UNSW Scientia Fellows and are global leaders in their areas of research.

  • The expected outcomes of the project are the development of a prototype tool (written in Python language) to identify the shoreline position from beach images. 
  • This tool will be compared to existing image processing techniques based on colour differences between sand and water. 
  • A journal publication with regards to the accuracy of this new technique to mapping shoreline change may also be warranted.